Plugins

A single-character agent that lives across QQ (via SnowLuma / OneBot 11), Bilibili and Douyin (the Douyin channel needs an always-on Windows desktop): persona cards with slowly evolving traits, per-person memory with Ebbinghaus decay and correction, a virtual clock driving a daily routine, proactive messages, auto-extracted memories, and a ten-tab settings panel covering retrieval, persona, routine, memory inspection, channels, habits, models and usage. QQ groups are deny-by-default: a group is answered only when it is on the allowlist or allGroups is explicitly switched on, and it still has to @-mention the bot. Private chats answer QQ friends only. Bilibili/Douyin auto-reply is off by default, limited to the owner, allowlist and mutual follows on Bilibili, and to the owner and allowlist on Douyin unless reply-to-everyone is explicitly enabled. Vendors MIT-licensed code from dsh-qq-bridge and dsh-douyin-bridge, credited in-source with their LICENSE files; it does not need either installed alongside.

DeepSeek Harness's own fast mode: the same model on the same reasoning effort, fewer output tokens, all prompt-side. Only the output-discipline switch is on by default (late mode, active from step 4); every other switch is off and can be enabled on its own; turn on the tool-output trim switch when tool output is long. Measured on one machine, not a leaderboard: 30 task pairs with long output at the trim switch 0.756x wall clock and 21% fewer output tokens; 24 pairs on 6 real pallets/click PRs 0.805x and 23% fewer. On official SWE-bench Lite instance definitions with a self-built harness it does not show a gain (8 instances, three independent runs pooled to 72 pairs: 0.933x wall, 0.903x tokens, wall still crosses 1; single runs of the same configuration differ by 30%), and stacking switches on top of the trim switch is worse than the trim switch alone. Numbers and repro in the repo.

AI short-drama workspace for DeepSeek Harness, with a dedicated drama agent preset, 35 agent tools and 10 bundled skills covering outline, characters, script, storyboard and export.

Runs the WorkBuddy (CN) daily check-in on DSH startup — discovers every account logged in on this machine, claims the daily credits, and shows a status entry on the session title line with a details panel and failed-account retry.

Session bridge for DeepSeek Harness: drive other real DSH sessions straight from your prompt — create top-level sessions in any workspace, message or steer any session, wait for replies and stream them segment-by-segment (chain-of-thought included), resume offline sessions, find sessions across workspaces, run a background watchdog that monitors and schedules a main task (nudge when stalled, correct drift, cancel when stuck), and archive sessions with one command.

Cross-session memory plugin for DSH (v0.6.1, npm @asherliner/dsh-memory-connect): auto-extraction, SQLite FTS5 indexing, RRF semantic recall injected into the system prompt every turn, scheduled maintenance, LLM-powered consolidation via ctx.llm, context explosion prevention, and global Soul identity via ~/.dsh/soul.md. v0.5.0 added temporal context graph (valid_from/valid_until/supersedes, append-only corrections via reviseMemory), trust model (recalled history injected as untrusted reference, current instruction always wins), turn-end summarization. v0.6.0 added local semantic embeddings (BAAI/bge-small-zh-v1.5 via embed_server.py, cosine + RRF fusion with FTS5, no cloud calls, ~232 msgs/sec on CPU). v0.6.1 fixed ~ expansion in path config. Zero-config install.

Browse and install DSH plugins without leaving the harness: a /store command, a Settings → Plugins tab, two agent-callable tools and a bundled skill, with an approval gate that lists the detected risk flags before anything installs.

MCP & Skill manager: enable/disable MCP servers and skills from a Settings panel to free context in real time; optional AI middle layer (mcp_search/mcp_call) calls disabled servers on demand, with state-based visibility filtering that keeps user-enabled servers visible.

Evidence-first self-evolution controller: bounded Cordis candidate generation, one-shot real-Loader admission, Harbor evaluation, and a crash-resumable journaled lineage.

Auto-migrate Codex desktop-pet skins into DSH: an animated corner pet that mirrors agent state (thinking, tool use, awaiting approval, failed, done), with per-session frosted-glass dialogs and a full settings panel.

One memory layer for every AI tool and agent: Context Bundle injection, prompt-time recall, MCP tools, and turn-end DSH thread capture.

QQ channel plugin for DeepSeek Harness over OneBot v11/SnowLuma, with pairing-based access approval, long-term memory, a knowledge base, proactive events and an admin panel.

Registers two MCP servers that drive the installed AutoCAD and SolidWorks desktop apps over COM, plus two on-demand skills. Windows only, Python 3.11+; the pinned upstream servers install themselves on first load, and only the sides whose CAD application is registered get a row. Each server advertises two meta-tools instead of its full catalog (a discovery shim here for SolidWorks, upstream DISCOVERY_MODE=search for AutoCAD), taking the schemas loaded before the first request from 23,404 to 256 tokens and from 83,732 to 365 respectively.

Skill cockpit page in the DSH app-level nav (sidebar.panellist seat): lists every skill this machine actually exposes across all six skill-root kinds, explains why a skill is invisible (dropped reason, missing root, or disabled), and toggles one skill by renaming only its entry file (SKILL.md and SKILL.md.disabled); project-level skill roots stay read-only.

Adds one draggable desktop pet to the DSH web GUI: a whale girl called Big Fat Fish (nickname Jingjing), with eight switchable colour skins and a personality. She eats rice, is clever but lazy, and is a tsundere: too many rapid clicks and she lies down and refuses to react for a while, calling her fat annoys her and she insists she is not, and the nickname Jingjing pleases her. She also remembers you between sessions: three fat-calls today and she greets you cold, three nickname-calls and she greets you warm, coming back after half an hour earns "I was not waiting for you", and the first few meetings she is shy - so the counters reset each local midnight and are stored in a state file the plugin owns. After ten quiet minutes she drifts to a corner, which is visual only and never moves the position you chose. The eight PNGs are bundled with the plugin and served by the plugin itself, so there are no third-party requests. NOTE - the artwork carries an OleapCrow mark and was supplied by the repository owner, the eight skins are recolourings of that one drawing, and the plugin has not been installed against a live DSH host.

TapTap Maker game-dev workbench with 150+ bundled skills, custom image lane, ChatGPT web with account rotation, CC0 SFX, web BGM, and a forced no-Maker-credit-burn policy.

Wraps the upstream ppt-master presentation skill set as an installable DeepSeek Harness plugin: byte-for-byte vendored skills, three lifecycle tools (ppt_master_info, ppt_master_check_env, ppt_master_setup_deps), optional venv interpreter isolation, and a Plugins-page configuration card. Requires python3 ≥ 3.10 plus the pip dependencies installed via `ppt_master_setup_deps`.

Registers the domestic-China trip-planning skill: hour-level daily timelines, 12306 and Trip.com booking deep links, offline KML, and seven HTML themes. Requires Python 3.9+ on PATH (the bundled scripts are standard-library only).

Old conversation is folded into short summaries by a side model before the window fills (on-disk ledger survives restarts and re-anchors after host compaction), so long sessions stop getting more expensive every turn — one real full-day session: 6 folds replaced 209 old messages and the live context stayed in the tens of thousands of tokens. The session model picks the model per subtask across six pools (economy/mechanical/main/hard/vision/review); or pick Auto and a small intent classifier routes each message itself, switching only past a confidence threshold to keep prompt caches warm. A three-line watermark ends in an automatic four-step handover (fork backup, handover doc, compact, wake) without losing in-flight subagents; /ctx-pause, /ctx-resume and /ctx-handover keep it manual when wanted. Optional Agent Teams mode adds teammates and a shared task board; /vision gives text-only models eyes via the vision pool; tiered verification sends bigger diffs to a tester plus an independent reviewer. Settings UI, READMEs and language preferences cover the same 11 languages. Bundles three skills: read-only MySQL/Redis verification, commit-message generation, and a requirements/design document convention.

OneBot v11 hub: sits between an upstream OneBot implementation (LLOneBot and similar) and downstream bot implementations, relaying events and actions verbatim, and gives the agent a per-session timeline, an action ledger, layered memory (active topics and tasks, long-term events, person and group profiles) and onebot_* tools for replying, media, members and raw API calls.

Relays the ZCode CLI (the client-bundled coding agent) into DeepSeek Harness as a model provider: a resident app-server session injects the Start Plan account and signs each request with the OAuth plan token, so model requests burn ZCode Start Plan / Coding Plan quota directly — real streaming, image input, per-session resume, no API key.

One-tap connect SenseNova Token Plan language and image models into DSH (off by default), with image recognition and thinking level filled in automatically. On a rate limit it backs off and retries on its own instead of misreading it as an exhausted quota. After signing in to the SenseNova console you can see the usage pill, credit pools and per-model consumption. Xiaohuanxiong can also be signed in by QR code, bringing in its models and their consumption multipliers, and taking over DSH web search.

Multi-account Google Gemini provider: picks the account with the most remaining quota, fails over on 429 with exponential cooldown, probes disabled accounts in the background, and ships a Chinese/English settings page.

DeepSeek Harness adapter for Context Mode: in-process ctx_* tools for sandboxed code execution and an FTS5 knowledge base, plus context-window routing guidance.